Local area networks take on many forms. A typical local area network (LAN) uses data packets to transfer data between nodes attached to the network. Using data packets to transfer data has proven to be an efficient means of allocating the available bandwidth of a network among a plurality of nodes. In most environments, the use of data packets to transfer data is an effective mechanism because of the nature of the data traffic carried by the LAN.
Data-packet-based network protocols are well suited to local area networks that service offices and wide area networks such as the Internet. This is because the data transferred between computing devices attached to these types of networks may be fragmented and delivered without worry as to the order of delivery or latency associated with propagating any given data packet through the network. So long as all of the data packets arrive at a destination node, they may be reassembled into coherent data.
Home networking is a newly emerging field. One objective in home networking is to allow various computing and entertainment devices to be connected to each other. Once these devices are connected to each other, informational data and entertainment content may be shared amongst various devices attached to the home network. Home networking users may want to transfer files between computers or share peripherals such as printers. In other situations, home networking users may want to access gateways to broadband service so they can share a single Internet connection amongst various devices. Home networking will most certainly be used for other services such as voice-over-IP (VoIP) and streaming media for entertainment.
Because the expectation of a household network user is so different than that of a typical office user, the nature of the data traffic pattern exhibited by a typical home networking structure is significantly different from other networking environments so far known. One such difference may be attributed to entertainment content flowing through the home networking structure.
Entertainment content must usually arrive at a destination node in proper sequence. In many situations, the latency between various streams of entertainment content must also be controlled. In some situations, some latency may be accommodated so long as the latency is deterministic and bounded by some reasonable maximum. Hence, it may not be appropriate to use data-packet-based network protocols in a home networking environment because of data fragmentation and the possibility that a data stream carrying entertainment content may arrive at the destination node either out-of-sequence or with excessive or non-deterministic latency.
The physical implementation of a home networking structure may also take on various forms. In some newer housing structures, Cat 5 or fiber cable may be available to connect computer and entertainment devices together. However, only the most modern houses may be pre-wired in this fashion. Because of this, the field of home networking has embraced alternative methods of connecting devices to each other. In some home networking systems, computer and entertainment devices may be connected to each other using power or telephone wiring. In yet other home networking implementations, a wireless communications scheme may be employed.
These three candidate transmission mediums, wireless, power wiring and phone wiring, each provide a much harsher environment for communication than cables designed specifically for networking (such as Cat 5). In particular, these three mediums are each subject to frequency selective impairment, wherein the signaling path between two nodes might support high-speed communications at some frequencies, but at other frequencies the ability to communicate might be poor or altogether lacking.
Most home networking systems rely on well-established, data-packet oriented transfer protocols. One such protocol is the well-known Transmission Control Protocol/Internet Protocol (TCP/IP). The TCP/IP protocol can use various physical communications medium. For instance, the TCP/IP protocol may be used in conjunction with Ethernet carried by copper cable. The TCP/IP protocol may also be used with a wireless physical layer. A power line based physical layer may also be used to support TCP/IP.
In networking terminology, the physical medium is controlled and accessed by a physical layer component. The physical, or “PHY” component may comprise hardware that enables higher levels of networking protocols to access the physical medium. The PHY component, which defines the method of attaching to the medium and the transfer of data packets, is typically augmented with a media access control (MAC) component. In some situations, the MAC component may be tightly coupled with the hardware that comprises the PHY component. The two may operate collectively to support access to the medium by higher levels of networking protocols.
Through evolution, the PHY/MAC components of various types of communications medium have been tailored to support transfer of fragmented data. And in some cases, the PHY/MAC components may actually exploit the fact that data may be fragmented into smaller data packets. One such exploitation is the notion of allowing collisions to occur when two or more devices attached to the medium attempt to transmit simultaneously. An example of this is the Carrier Sense, Multiple Access—Collision Detection (CSMA/CD) physical access method used in Ethernet PHY/MAC components. Because data is fragmented into smaller data packets, it is easier to tolerate retransmission of data packets lost as a result of collisions.
Home networking systems that use a CSMA/CD access method may work just fine if the data traffic pattern carried by the network were to more closely resemble that of an office environment. The problem is that a home networking system must be capable of carrying not only informational data, but entertainment content as well. Because entertainment content should be delivered in proper sequence and with controlled latency, CSMA/CD PHY access may be an inappropriate mechanism to impart the entertainment content onto the physical networking medium.
An isochronous channel may be better suited for transferring entertainment content from one node in a home networking environment to another. By definition, an isochronous channel ensures sequential delivery of a data stream at a controlled rate, and with deterministic latency. Most home networking systems cannot support isochronous transfer because they rely on traditional data-packet oriented transfer protocols and their PHY/MAC access methods.
Many known home networking systems use the power lines that service convenience outlets in a home as the physical medium between data and entertainment nodes. This is because power outlets are usually found everywhere someone might want to use a networked device. In one known power-line based system, the PHY component uses orthogonal frequency division multiplexing (OFDM) as the basic transmission technique. In one system, the OFDM access method is used on a transient basis to support varying levels of communications bandwidth.
Some home networking systems that use OFDM media access also utilize a MAC protocol that is a variant of the well-known carrier sense multiple access with collision avoidance (CSMA/CA) protocol. The CSMA/CA protocol is typically used in wireless LAN systems and relies on the use of a jamming signal that is used to detect collisions. Several variations to the classic CSMA/CA protocol may be found in some MAC components. The most common enhancements provide support for priority classes, fair access, and controlled latency. The use of CSMA/CA means the PHY must support burst transmission and reception; that is, each client enables its transmitter only when it needs to send data.
OFDM divides a high-speed data stream, such as entertainment content, into a plurality of parallel bit streams, each of which has a relatively low bit rate. Each individual bit stream may then be used to modulate one of a series of closely spaced carriers. The carriers are spaced so that they do not interfere with each other and the spacing between carriers is dictated by the bit rate of the modulating data.
Many different modulation techniques may be applied independently to the individual carriers. In some systems that use power lines as the communications medium, the quality of the power line as a communications medium may be assessed and a modulation technique may be selected for each carrier based on a realistically sustainable modulation density at the carrier frequency over the medium between two nodes.
In some power line based home networking architectures, the same transmitting node may not necessarily use all of the available carriers simultaneously. This means that a plurality of devices attached to the network medium, i.e. to the household electrical or phone wiring, may contend for the same communications resources and that some or all of these contenders will be able to acquire some portion of those resources. In these PHY/MAC embodiments, a plurality of network devices may simultaneously transmit data.
Home networking structures that incorporate PHY/MAC components that comprise OFDM media access with enhanced CSMA/CA protocol still continue to operate under the traditional packetized data paradigm. As a result, entertainment data streams may not have continuous or deterministic access to the communications medium. After every packet is sent, network bandwidth used for entertainment streams must be relinquished. Once the bandwidth is released, it may not be available for a subsequent data packet because of other devices attempting to transmit their data packets. This may result in additional latency because each subsequent attempt to continue transmission of the entertainment data stream will require a contest for available communications resources. The number of carriers available to convey data may vary over time as other nodes on the network capture resources or portions of the available communications spectrum suffer temporal impairments.
Local area networks take on many forms. A typical local area network (LAN) uses data packets to transfer data between nodes attached to the network. Using data packets to transfer data has often proven to be an efficient means of allocating the available bandwidth of a network among a plurality of nodes. In many network environments, the use of data packets to transfer data is an effective mechanism because of the nature of the data traffic carried by the LAN.
Packet-based network protocols are well suited to local area data networks that service offices and to wide area networks such as the Internet. Packet protocols are very useful for transfer of computer data in such networks for several reasons. First, the bandwidth demand for computer data tends to be highly variable and unpredictable, and moreover there is flexibility to trade off bandwidth against delay in delivering the data. Second, packet protocols are very useful in large networks where data can reach its destination via multiple paths. In a packet system, each packet can be routed via a path that has the most available bandwidth at the time of routing. This routing efficiency makes packet delivery desirable even for fixed bandwidth services such as telephony. Data transferred between computing devices attached to these types of networks may be fragmented into packets and delivered successfully even when packets arrive out of order or are subject to substantial latency associated with propagating any given data packet through the network. So long as all of the data packets arrive at a destination node, they may be reassembled into coherent data. Office and wide area networks typically are not used for applications with extreme sensitivity to the delay encountered by the data being transmitted. Moreover, the medium that carries the network usually has ample bandwidth for the application, and if the bandwidth does become constrained, the medium can be expanded to provide more by a variety of techniques.
To summarize, there are four key attributes of packet systems in relation to the office environment that should be considered with respect to the use of such systems for home networking. These attributes are:                1. The bandwidth required to support the intended applications varies rapidly over time.        2. The intended applications are relatively insensitive to substantial latency in delivery of the data.        3. There is substantial efficiency to be gained in dynamic routing of packets via the best of many possible paths.        4. Bandwidth on the physical medium is relatively inexpensive, and more can be added at relatively low cost if needed.        
Home networking is a newly emerging field. Early home networks have been used to provide computer-oriented capabilities in the home similar to those that users have become accustomed to at work. These include file sharing, peripheral sharing, and sharing access to the Internet via a gateway. But in the next generation of home networking, bandwidth requirements will be dominated by entertainment applications, and these applications have very different requirements and behavior compared to computer data.
Because the applications for a household network are so different from those of a typical office network, the nature of the data traffic pattern exhibited by a typical home networking structure is significantly different as well. By comparison to the four key attributes of the office environment, a home network designed to carry entertainment traffic has the following characteristics:                1. The majority of the traffic in the home network has a fixed bit rate.        2. The majority of the traffic in the home network is intolerant of large latency.        3. There is only one routing path, so the efficiency gains in packet routing do not exist.        4. Bandwidth is a scarce resource with a definite limit.        
Entertainment content is almost always transmitted at a fixed bit rate. A standard definition digital television channel might require a 6 Mbps connection. A high definition digital television channel might require 24 Mbps. A CD audio connection requires about 1.5 Mbps. These requirements do not change for the duration of the application. This is in stark contrast to a computer application like web surfing, where there is no bandwidth requirement while the user is reading a web page, and then suddenly a high demand when he clicks on a link.
Entertainment content is also typically very sensitive to delay. If a video frame is not delivered in time, the video display has no information to show, and the screen must either be blanked or the display unit must rely on some other algorithm to estimate the missing data. Either of these events cause undesirable artifacts in the image displayed. In addition, audio and video must be synchronized in time so that sound and motion appear coordinated, requiring that the delay difference between the audio and video information carried on the network must be controlled. Networks that carry video and audio data typically include elastic buffers (memory) that smooth out some degree of variation in delivery time, but to tolerate long delays, the memory needed is very large, and the cost of this memory is prohibitive. What is more, the total delay allowed must still be small enough so that the system seems responsive when the user changes channels.
This extreme sensitivity to latency suggests that it may not be appropriate to use packet-based network protocols in a home networking environment because these protocols typically have potentially large and unpredictable latencies.
The physical implementation of a home networking structure may also take on various forms. In some newer housing structures, Cat 5 or other network cable may be available to connect computer and entertainment devices together. However, only the most modern houses may be pre-wired in this fashion. Because of this, the field of home networking has embraced alternative methods of connecting devices to each other. In some home networking systems, computer and entertainment devices may be connected to each other using power or telephone wiring. In yet other home networking implementations, a wireless communications scheme may be employed.
These three candidate transmission mediums, wireless, power wiring and phone wiring, each provide a much harsher environment for communication than cables designed specifically for networking (such as Cat 5). In particular, these three mediums are each subject to frequency selective impairment, wherein the signaling path between two nodes might support high-speed communications at some frequencies, but at other frequencies the ability to communicate might be poor or altogether lacking.
Regardless of the quality of the path, there is only one route available for the data to take on any of these three types of medium. Thus a packet based protocol does not accrue any advantage from routing efficiency.
In each of the three of the potential media for home networking, there is a limit to how much bandwidth is available. In each case the limit results from regulation on spectrum usage. For a power line system, the usable bandwidth lies below 30 MHz, and there are many segments of this band that must be avoided to prevent interference with other licensed radio services. The net result is that there is on the order of 23 MHz of useful bandwidth. A home network that supports entertainment must be able to provide about 50 Mbps of aggregate bandwidth to the application layer while operating confined to this band. This requires a much greater bandwidth efficiency than any current office network can achieve.
Most home networking systems rely on well-established, data-packet oriented network layer protocols. One such protocol is the well-known Transmission Control Protocol/Internet Protocol (TCP/IP). The TCP/IP protocol can use various physical communications medium. For instance, the TCP/IP protocol may be used in conjunction with Ethernet carried by copper cable. The TCP/IP protocol may also be used with a wireless physical layer. A power line based physical layer may also be used to support TCP/IP.
TCP/IP is a general-purpose protocol that can be used on many networks, but to use it on different media such as power line, phone line or wireless requires that it interface to lower layer protocols designed specifically for these media. These lower layer protocols typically consist of two component protocols, a physical layer (PHY) protocol that determines the modulation, coding and signal format to be used to transmit information on the wire, and a medium access control (MAC) protocol that controls how the medium is accessed and shared by the various network nodes. In some cases the MAC may also provide capability to detect and correct transmission errors. The MAC component may be tightly coupled with the hardware that comprises the PHY component. The two layers operate collectively to support access to the medium by higher levels of networking protocols.
Through evolution of networking technology, the PHY/MAC components of many types of communications networks have been tailored to support transfer of packet data. A common approach to transmission of packet data is the Carrier Sense, Multiple Access—Collision Detection (CSMA/CD) physical access method used in Ethernet PHY/MAC components. In this scheme, a set of rules govern the time at which devices may attempt transmission with the intent of reducing, but not eliminating, the chance of a packet collision—that is, the event where two or more nodes attempt to transmit at the same time. When a collision occurs, the data transmitted is typically not received by the intended destination. Packets may also be lost due to impairments that degrade the channel. Lost packets are typically recovered by having the destination acknowledge packets that are received. The source then retransmits packets that are not acknowledged.
Because transmission times are not coordinated among the network nodes, collision events happen at unpredictable times and the amount of time expended in recovering lost packets is also unpredictable. Moreover, even when there is no collision, a node may have to wait an unpredictable amount of time for the medium to be released by another node, and thus encounters a variable delay.
The CSMA/CD method works very well for computer data where variability in delay is tolerable. It is not uncommon for delays in LANs for computer data to be as low as a few microseconds or as large as half a second. The user of the network typically doesn't notice these short-term fluctuations in delay as long as the overall throughput is good. The protocol is effective at providing reasonably good long-term throughputs, and it is very simple and cost effective to implement. But a CSMA/CD protocol typically wastes a substantial percentage of the available bandwidth as nodes wait to access the channel for various amounts of time chosen to minimize the chances of collision.
Home networking systems that use a CSMA/CD access method may work well if the data traffic pattern carried by the network resembles that of an office environment. But a home networking system must be capable of carrying not only computer data, but entertainment content as well. Because entertainment content must be delivered with controlled latency, CSMA/CD PHY access may not be the best choice for a medium access control protocol.
In addition to these latency and efficiency problems, it is difficult to use a CSMA protocol with bandwidth efficient modulations such as high order quadrature amplitude modulation (QAM). Use of QAM can allow transmission rates of many bits per second per Hz—prior art systems such as ADSL use QAM to achieve as much as 10 bits per second per Hz or more. However, high order QAM modulations require that the modem receiver must track the correct sampling time, carrier phase, and signal amplitude with extremely good accuracy. Obtaining this accuracy typically requires a long initial training time. But in a packet system where the source of the packet is not known a priori, the receiver must train on each packet received individually without use of knowledge derived from prior receptions. To support high order QAM, each packet would thus have to include a long training sequence (that is, a known signal that can be used to form an estimate of timing, phase and amplitude). The overhead introduced by this training sequence typically consumes most of the bandwidth savings achieved by using the higher order modulation, so that there is no improvement to be gained.
ADSL is able to use high order QAM because it uses a circuit type of connection. The modem acquires timing, phase and amplitude estimates during a training period when the connection is first established and then continues to track these parameters throughout the data transmission. In home networking with entertainment content, establishing circuit connections can also provide the ability to support high order QAM modulation. But unlike ADSL where the medium is used to connect only two devices, the network medium must connect multiple devices and thus must support the establishment of multiple circuit connections simultaneously. This can be achieved through well known schemes such as Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Code Division Multiple Access (CDMA) or combinations of all three.
Historically, the term “circuit” meant that there was a single wire dedicated to carrying the data to be transferred between two nodes and that the capacity of this wire was unavailable to any other nodes while the transfer was in progress. More recently, the word “circuit” has come to describe a connection that provides a known and constant amount of data carrying capacity in a way that appears to provide continuous availability of the channel to the applications that are exchanging data. For example, a “virtual circuit” can be provided over a medium in which time is divided into intervals and a fixed slot of time and bandwidth in each interval is provided to the applications. Other applications can be allocated other time slots and/or bandwidth so that multiple virtual circuits can be supported simultaneously by a single wire. The key attribute of a circuit connection as the term is used herein is that the connection provides the application with a constant throughput capability that does not vary as other data traffic is added or removed from the network, and that this is done by reserving access to the channel at appropriate times and frequencies for the exclusive use of the nodes using the circuit.
When OFDM is used on a circuit connection, even if the circuit connection is not continuous in time, the fact that the times at which the signal will be present are predictable allows the receiver to use information about the signal obtained during previous transmissions (such as the appropriate receiver gain, symbol timing, carrier phase and frequency) to assist in demodulation of subsequent transmissions. This vastly improves the ability to support denser modulations and thus provide more efficient use of the available bandwidth.
Home networking systems that use the power wiring in a home as the physical medium between data nodes are currently available on the consumer market. This approach to home networking is attractive because power outlets are is usually found virtually everywhere someone might want to use a networked device. In one known power-line based system developed by the HomePlug PowerLine Alliance consortium, the PHY component uses orthogonal frequency division multiplexing (OFDM) as the basic transmission technique. In this system, the OFDM signaling is used in combination with a packet based MAC protocol that is a variant of the well-known carrier sense multiple access with collision avoidance (CSMA/CA) protocol. Several variations to the classic CSMA/CA protocol may be found in some MAC components. The most common enhancements provide support for priority classes, fair access, and controlled latency. The use of CSMA/CA means the PHY must support burst transmission and reception; that is, each client transmits its data in limited duration transmissions and then frees the channel for contention by other nodes. OFDM divides a high-speed data stream, into a plurality of parallel bit streams, each of which has a relatively low bit rate. Each individual bit stream may then be used to modulate one of a series of closely spaced carriers. The bits in these individual streams are formed into symbols, with the number of bits in a symbol being the same for each symbol transmitted by a particular carrier, but not necessarily the same for different carriers. The symbol rate for all the carriers is the same, but the bit rate supported by each carrier is the product of the symbol rate and the number of bits per symbol, and thus may be different for each carrier. The number of bits per symbol is called the modulation density and is a function of the quality of the channel at each carrier frequency. Generally speaking, better channel quality allows the use of a greater number of bits per symbol and thus allows greater efficiency in bandwidth utilization. Channel quality is determined by a variety of factors, including, but not limited to, signal-to-noise ratio and distortion due to the channel response.
The carriers are spaced so that they do not interfere with each other and the spacing between carriers is dictated by the symbol rate. This approach to modulation is advantageous for channels where the degradation encountered by the signal varies substantially as a function of frequency, as is the case for the wireless, phone line, or power line channels. The modem can transmit data at a higher bit rate on carriers in frequency bands that have better quality by using denser modulation constellations. Thus the network can be rate adaptive, matching the overall bit rate to the quality of the channel.
Many different modulation techniques may be applied independently to the individual carriers. In some systems that use power lines as the communications medium, the quality of the power line as a communications medium may be assessed and a modulation technique may be selected for each carrier based on a realistically sustainable modulation density at the carrier frequency over the medium between two nodes.
In one existing power line based home networking architecture (HomePlug v1.0), if the channel is unusable with any modulation density for transmission between two nodes at the frequencies occupied by certain carriers, these carriers are not used to bear data. In this example architecture, while the two nodes are communicating no other nodes may make use of these unused carriers, even if the channel at those frequencies is capable of supporting transmission between these other nodes. This causes a substantial inefficiency in the usage of the available bandwidth and reduces the capacity of the network from what might otherwise be possible.
Home networking structures that incorporate PHY/MAC components that comprise OFDM media access with enhanced CSMA/CA protocol still continue to operate under the traditional packet data paradigm. As a result, in these systems entertainment data streams do not have continuous or deterministic access to the communications medium. After every packet is sent, network bandwidth used for entertainment streams must be relinquished. Once the bandwidth is released, it may not be available for a subsequent data packet because of other devices attempting to transmit their data packets. This may result in additional latency because each subsequent attempt to continue transmission of the entertainment data stream will require a contest for available communications resources. The number of carriers available to convey data may vary over time as other nodes on the network capture resources or portions of the available communications spectrum suffer temporal impairments.
Existing packet based technologies for home networking using either wireless, phone line or power line as the medium all suffer from two major shortcomings: first, the efficiency of bandwidth utilization is so poor that the available bandwidth is insufficient to support transport of high bit rate entertainment content such as high definition TV signals, and second the latency of delivery of data is inadequate. These two drawbacks make existing systems ineffective for distributing entertainment content.
What is needed then is an approach to networking that is specifically designed with the intent of carrying entertainment content and that specifically addresses the issues of bandwidth utilization and latency. The invention described herein does this.